On 5/2/19 11:43 PM, Thomas Huth wrote: > On 03/05/2019 00.02, Eric Blake wrote: >> On 4/28/19 10:21 AM, Thomas Huth wrote: >>> QEMU iotest 221 is failing for me, too, when I run it with -raw: >> >> Which filesystem? > > ext4 again. > > $ stat -f /home/thuth/tmp/qemu-build/tests/qemu-iotests/ > File: "/home/thuth/tmp/qemu-build/tests/qemu-iotests/" > ID: 1e68b4a412e09716 Namelen: 255 Type: ext2/ext3 > Block size: 1024 Fundamental block size: 1024
Odd that it is so small; these days, most ext4 systems have a block size
of 4k.
>
> Maybe the "check" script should report the output of "stat -f", too?
Wouldn't hurt, although that doesn't tell us all of the file system
tuning parameters that might be important to reproducing a problem.
>>> +++ /home/thuth/tmp/qemu-build/tests/qemu-iotests/221.out.bad
>>> 2019-04-28 17:18:52.000000000 +0200
>>> @@ -7,10 +7,10 @@
>>> [{ "start": 0, "length": 43520, "depth": 0, "zero": true, "data": false,
>>> "offset": OFFSET}]
>>> wrote 1/1 bytes at offset 43008
>>> 1 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>>> -[{ "start": 0, "length": 40960, "depth": 0, "zero": true, "data": false,
>>> "offset": OFFSET},
>>> -{ "start": 40960, "length": 2049, "depth": 0, "zero": false, "data": true,
>>> "offset": OFFSET},
>>> +[{ "start": 0, "length": 43008, "depth": 0, "zero": true, "data": false,
>>> "offset": OFFSET},
>>> +{ "start": 43008, "length": 1, "depth": 0, "zero": false, "data": true,
>>> "offset": OFFSET},
>>
>> Ugh. Hole granularities are file-system specific, so we need to figure
>> out some way to fuzz the input. It might also be possible to pick nicer
>> size numbers - perhaps if the test image is sized at 64k+1 instead of
>> 43009 (84*512, but NOT evenly divisible by 4k), the +1 byte is more
>> likely to be directly one a hole boundary, rather than being somewhere
>> that causes rounding the hole boundary 2k earlier because of 4k or 64k
>> sizing constraints.
>
> Ok ... sounds like that's definitely something I'd like to leave to you
> or one of the block guys to fix.
I can certainly prepare a patch that widens the file to 64k+1 instead of
43008+1, but since I can't (yet) reproduce the failure, I'd be relying
on you to verify that it makes a difference.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3226
Virtualization: qemu.org | libvirt.org
signature.asc
Description: OpenPGP digital signature
